A Preliminary Investigation Into the Use of the Emotional Contagion Effect in the Exercise Environment

The term “emotional contagion” (EC) refers to the tendency for humans to mimic the emotions of others; this can occur either consciously or importantly for this investigation, unconsciously (Hsee, Hatfield, Carlson, & Chemtob, 1990; Neumann & Strack, 2000; Wild, Erb, & Bartels, 2001). The EC effect has been shown to manifest in many situations and in response to a variety of transfer sources. Using speech as the emotional transfer medium, Neumann and Strack (2000) asked participants to listen to either a “neutral” or emotionally charged (“happy” or “sad”) reading of a philosophical text. Upon hearing the text, participants were first asked to rate their present mood state (“How do you feel right now?”). After this, they were presented with a list of emotional adjectives (e.g., happy, cheerful, sad) and asked to rate how strongly they were feeling each. All responses were recorded on a 10-point Likert scale. Although there was no change in the specific, adjective-based feeling states, the emotional tone of the reading did induce the congruent mood state in the listeners.

Facial expression provides a particularly potent form of emotional transfer (Dimberg & Ohman, 1996; Hess & Blairy, 2001). Applying this medium, Hess and Blairy (2001) presented participants with short video clips of persons displaying various emotional expressions (anger, sadness, disgust, and happiness). Facial Electromyography (EMG) activity and self-reported ratings of subjective affect were assessed. Even when relatively weak emotional expressions were used, the EMG recordings showed differentiated patterns of facial muscle activity in response to the stimuli, and these activation differences were found to correspond to the emotion being presented. Affective state was also found to correspond to the emotional stimuli. Unfortunately, insufficient data were presented to allow an effect size to be calculated. Wild et al. (2001) also used emotionally loaded facial expressions to study the EC effect. In this instance, static images were taken from the Pictures of Facial Affect Compendium (Ekman & Friesen, 1976) and digitally morphed to produce numerous emotions (happy and sad) at varying intensities (75%–150%). Using a 0 to 100 linear scale, the pictures were rated on how much they evoked associated feeling states. The tone of the photographs did have a significant effect on participants. For example, it was found that the “happy” and “sad” pictures provoked the corresponding feeling state. It was also shown that emotions were transferred relatively quickly and the stronger the expression the stronger the effect. Again, effect sizes were incalculable from the data presented. The inducement of EC through pictures of static facial images has also been observed by others; for example, Dimberg and Thunberg (1998) and Dimberg, Thunberg and Elmehed (2000). While the preceding examples offer support for the contagious nature of emotions, evidence from outside controlled laboratory settings would be more indicative of any potential practical application.

Some researchers have studied the effect in what might be termed “real-world” environments. Johnson (2008) examined EC and leader–follower (employee) relationships in the education profession. A sample of teachers was asked to complete various measures of affect as well as a charismatic leadership questionnaire. It was hypothesized that “follower” affect would be positively related to perceived charismatic leadership qualities, namely, attributed charisma, idealized influence, and inspiration motivation. Correlational analysis provided some support for the author’s assertions with both “affect at work” (r = .34, d = 0.72) and “general” affectivity (r = .37, d = 0.79) being significantly associated with perceived leadership charisma. While significant, it must be recognized that the correlations were not particularly strong. In this instance, the effect sizes offer better evidence of an EC effect. Analyzing the role of EC in customer relations, Hennig-Thurau, Groth, Paul, and Gremler (2006) contrived a situation to assess the role of smiling on interpersonal interactions between employees and customers. Actor “employees” were asked to display either “high” or “low” smiling behavior when dealing with customers; this was also subcategorized into “high” or “low” authenticity. Customer affect was measured via a number of 7-point Likert items prior to, and after encountering the employee. Overall, there was no statistical relationship between level of employees’ smiling and customers’ affective responses. However, evidence indicated that that the perceived authenticity affected the outcomes. Specifically, when the employees engaged in deep acting, evidence of an EC effect did emerge (partial η²= 0.14, d = 0.79).

EC has also been studied within authentic sport environments. For example, Moll, Jordet, and Pepping (2010) investigated how the penalty goal celebrations of international footballers affected other on-field players. More than 300 video clips of penalty kicks were assessed for their (physical) emotional content. Phi correlations revealed a significant association between certain celebratory behavior and competition outcomes. It was found that the team whose members displayed triumphal gestures (e.g., expanded chest and raised arms) after a successful attempt were more likely to win the penalty competition (φ² range = 0.16–0.29, d = 0.32–0.60). The authors surmised that this was due to emotional convergence. Specifically, because such expressions are synonymous with pride and confidence, a more positive mindset was induced in team members. Conversely, the opposing teams’ players were believed to have perceived such actions as dominating and as such, negative emotions were experienced. Totterdell (2000) assessed the effect of “collective” mood among professional cricketers. Players from four teams were supplied with pocket computers and asked to record their own moods and the perceived mood of the team as a whole. Pooled time-series regression analysis revealed significant, and positive correlations between the mood of the team and that of the individual players (r = .64, d = 1.66); for example, the happier the collective, the happier the individual. Totterdell attributed the outcomes to what he termed mood linkage—in other words, EC.

The preceding examples support the notion that emotions can be transferred to individuals via emotionally charged stimuli. With its potential for mood enhancement, it is possible that the EC effect could offer benefits to exercise participants. Affective state is considered to be an important determinant of exercise adherence (Williams, 2008). The suggestion is that positive experiences engender positive feelings toward a behavior, thereby increasing the likelihood it will be repeated (Ekkekakis, Parfitt, & Petruzzello, 2011; Godin, 1994; Williams, 2008). In essence, positive feelings act as reinforcements (Annesi, 2002). There is empirical support for this view. For example, using Structural Modelling Analysis, McAuley, Jerome, Marquez, Elavsky, and Blissmer (2003) found that for older adults, affect experienced during exercise had a direct effect on self-efficacy and ultimately program adherence (β = 0.26). Annesi (2002) also showed post-exercise feeling states to be correlated with adherence. In this case, however, the association was only applicable to individuals identified as possessing low self-motivation (r = .48, d = 1.09). Positive affect is also thought to be related to what some consider to be one of the most important determinants of exercise adherence, namely enjoyment (Raedeke, 2007; Hagberg, Lindahl, Nyberg, & Hellenius, 2009). Miller, Bartholomew, and Springer (2005) suggest that affective state might be fully mediated by this construct.

Another potential determinant of the exercise experience is effort sense, commonly referred to as rate of perceived exertion (RPE). For example, Annesi (2002) reported an interaction between effort sense and enjoyment for individuals classified as possessing a low motivation to exercise. Specifically, as perceived physical exhaustion increased, a decrease in positive engagement, a construct that relates to feelings of enthusiasm and happiness, was observed. Perception of effort also appears to be related to affective state. Hardy and Rejeski (1989) reported that a negative, linear relationship exists between these two constructs (r = −.56, d = 1.35). Additionally, O’Halloran, Murphey, and Webster (2005) found that reductions in in-task RPE accounted for between 7% and 9% of mood improvement during exercise. While these examples provide evidence of a link between RPE and affect, it is currently unclear as to whether the relationship is reciprocal. That is, whether improvements in affect lead to reductions in effort sense. It is possible given that psychological factors such as self-efficacy (Rudolph & McAuley, 1996), social desirability (Coquart, Dufour, Groslambert, Matran, & Garcin, 2012), anxiety, and depression (Watt & Grove, 1993) have been shown to mediate RPE. Ultimately, an ability to manipulate effort sense might influence adherence. Utilizing the Physical Exhaustion subscale of the Exercise Induced Feeling Inventory (Gauvin & Rejeski, 1993), Annesi showed that among individuals with low motivation to exercise, perception of effort was negatively associated with program adherence (r = .62, d = 1.58).

In summary, EC has been shown to manifest in many situations and via numerous emotional transfer mechanisms. The fact that emotions can be induced through this phenomenon might have practical application for those involved in exercise. There is evidence to suggest that factors such as affect, enjoyment, and effort sense mediate the exercise experience. Therefore, if the EC effect can induce a more positive exercise experience, then it could have implications for program adherence. To date, no study has attempted to utilize the EC effect in this environment. This investigation examined the effect of positively valenced facial imagery upon ratings of affect, enjoyment, REP, and perceived exercise experience. Participants performed two aerobic exercise sessions, one without any EC stimuli (baseline) and another under one of three conditions (control, neutral, or happy). The EC Scale was also utilized to determine whether the participants were susceptible to the EC effect.

Hypothesis. Positively valenced facial imagery will invoke increases in ratings of affect, enjoyment, subjective exercise experience, and lower session RPE.

Method

Analysis

A 3 (condition) × 2 (trials) Factorial Multivariate Analysis of Variance (MANOVA) with Bonferroni correction factor applied was used to assess for main effects and condition interactions. To uncover any condition × trial effects, dependent t tests with p set at .016 were utilized. Effect size, mean differences, and 90% Confidence Intervals (CI) were also calculated (Table 1). Correlational analysis was used to assess for relationships between level of emotional susceptibility and experimental DVs. Based on the assumption that a positive relationship should exist, a one-tailed analysis was implemented.

OPEN IN VIEWER

Table 1.

Descriptive Statistics, Effect Sizes, Mean Difference and Ranges for Each Dimension of the Emotional Contagion Scale.

Subscale	M	SD	Range
Happiness	5.68	1.28	3.00–8.00
Love	6.87	1.92	3.00–12.00
Fear	8.17	1.93	3.00–12.00
Anger	8.17	2.18	4.00–12.00
Sadness	7.95	1.71	3.00–11.00

Results

EC ratings ranged from 3 to 12 for each subscale. The median value for each subscale was 7.50 (low scores were indicative of high EC susceptibility). The values (Table 1) suggested that for the most important emotion “happiness,” the participants were prone to EC.

The mean Session RPE data confirmed that the exercise intensity was low–moderate (American College of Sports Medicine, 2006).

For the factorial MANOVA, Box’s Test of Equality was significant (p = .02), so Wilks’ Lambda was used for the analysis. There was no significant main effect between Trial 1 and Trial 2, F(7, 38) = 2.01, p = .08. For condition interactions, Wilks’ Lambda revealed no significant interactions for any of the variables, F(14, 78) = 1.10, p = .37. There were no significant interactions with condition: Session RPE, F(2, 44) = 1.02, p = .36; Feeling Scale, F(2, 44) = 2.27, p = .11; Enjoyment, F(2, 44) = .64, p = .53; Positive Well-Being, F(2, 44) = .01, p = .99; Psychological Distress, F(2, 44) = 1.63, p = .20; and Fatigue, F(2, 44) = .21, p = .80 (Table 2).

OPEN IN VIEWER

Table 2.

Descriptive Statistics, Effect Sizes, Mean Difference and 90% Confidence Intervals for Each Condition.

Condition and scale	Trial 1		Trial 2		Cohen’s da	M Diff.	90% CI
	M	SD	M	SD
Happy
Session RPE	2.43	1.09	2.06	0.77	−0.56	0.37	0.05, 0.69
Session FS	1.68	1.53	2.25	1.29	0.47	0.57	.001, 1.13
Enjoyment	4.50	1.71	3.93	1.28	−0.61	0.57	0.11, 1.01
PWB	16.37	3.82	15.56	2.47	−0.20	0.81	0.79, 2.24
PD	5.93	2.31	5.56	2.47	−0.11	0.37	−0.69, 1.44
Fatigue	8.43	5.92	7.12	3.72	−0.21	1.31	−1.01, 3.64
Neutral
Session RPE	2.84	0.68	2.94	1.17	0.06	0.10	10.58, 0.37
Session FS	2.47	1.57	1.88	1.99	−0.32	0.59	−0.22, 1.37
Enjoyment	5.05	1.12	4.78	1.08	−0.29	0.27	−0.13, 0.65
PWB	19.05	4.57	18.21	4.77	−0.28	0.84	−0.67, 2.31
PD	5.25	1.66	6.52	4.77	0.31	1.27	−2.92, 0.39
Fatigue	8.26	4.53	7.57	4.45	−0.14	0.69	−0.12, 2.58
Control
Session RPE	2.50	1.24	2.16	0.71	−0.27	0.34	−0.34, 1.00
Session FS	2.91	1.31	2.66	1.37	−0.20	0.25	−0.33, 0.84
Enjoyment	5.25	1.35	5.03	1.44	−0.20	0.22	−0.31, 0.65
PWB	19.08	5.24	18.33	6.61	−0.24	0.75	−0.81, 2.32
PD	4.75	1.76	4.33	1.15	−0.36	0.42	−0.18, 1.02
Fatigue	5.56	2.57	5.50	2.02	−0.01	0.06	−1.45, 1.78

The correlation between the “happiness” dimension of the EC scale and the Positive Well-Being subscale of the SEES was significant, r = −.24, p = .04. None of the remaining correlations with EC were significant: Session RPE, r = .12, p = .19; Feeling Scale, r = −.14, p = .14; Enjoyment, r = −.10, p = .23; Psychological Distress, r = −.15, p = .14; and Fatigue, r = .00, p = .48.

Discussion

Research has shown that emotions can be transferred to individuals through a process known as EC. It was argued that if the EC effect could be used to influence emotional valence positively and reduce the sense of effort, then the phenomena could provide a simple means of enhancing the exercise experience and ultimately program adherence (e.g., Annesi, 2002; Godin, 1994; Raedeke, 2007; Williams 2008). Despite the EC effect being well documented in other situations, there was little evidence of it in this investigation, despite the fact that participants were susceptible to the Happy dimension of emotional transfer. The primary statistical analysis (factorial MANOVA) revealed no significant effects for any of the dependent variables. The supplementary analyses (effect size, mean difference, and 90% CI) did offer some evidence of EC. In the Happy condition, there were moderate effect sizes for Session RPE and affect and the 90% CIs were in the hypothesized direction. Importantly, the effect size for affect (d = 0.57) was comparable to those presented by Hennig-Thurau et al. (2006) and Johnson (2008). Furthermore, it was only in this condition that a positive change in these variables was observed.

The outcomes need to be interpreted with caution, as there was also a moderate negative effect for Enjoyment (d = −0.61). Given the supposed interaction between these variables, a corresponding increase in Enjoyment might have been expected. The apparent effect for RPE and affect had no corresponding affect upon subjective exercise experience. It is, therefore, difficult to accept the effect sizes as being indicative of a “real,” practical effect. As an additional measure, correlational analysis was conducted between the Happy dimension of the EC scale and dependent variables. These data provide some indication as to whether emotional susceptibility was an influencing factor. A significant, yet weak, correlation existed between Happy ratings and Personal Well-Being; however, no other associations were evident. So, despite some encouraging effect sizes for two of the variables, it must be concluded that the data do not support the use of positive emotional imagery within the exercise environment. With these outcomes in mind, we will explore some reasons why the EC effect was not observed. It is hoped that this will aid future research in this area.

One possible explanation for lack of an EC effect is that the participants did not actually see the stimuli. In studies that have also used facial expressions (e.g., Hess & Blairy, 2001; Wild et al., 2001), participants were instructed to look directly at the images. In this investigation, the authors went to great lengths to conceal the nature of the experiment and thus did not take this course of action. It was assumed, because of the exercise mode and the fact that the images were sizeable and placed directly in front of the exercise group, that the stimuli would be readily observed. Therefore, the authors consider this explanation unlikely. The EC effect has been shown to manifest within seconds (Hess & Blairy, 2001; Wild et al., 2001) and so even cursory glances should have been sufficient to alter psychological state of the exercisers. Also, according to Bayle, Henoff, and Korak-Salmon (2009), stimuli, including facial expressions, are readily processed via the peripheral vision system. Nevertheless, perhaps researchers need to make concerted efforts to direct participants’ attention toward any stimuli, or use an exercise mode (e.g., treadmill running) in which stimuli can be placed in the central visual field. As has been argued, implementing such strategies could compromise both internal (demand characteristics) and external (ecological) validity.

Another aspect that requires consideration is temporality. It is unclear how long changes generated by EC persist. If transient in nature, then it might be that the stimuli were observed and emotional state did change but was modulated throughout the exercise session. Retrospective measures like those used in this investigation might lack the sensitivity to discover such momentary changes. Future research might consider utilizing in-task measures at regular intervals throughout an exercise session; this could be performed in conjunction with any attempts to direct participants’ attention toward the stimuli. However, even if this did provide evidence of EC, it could be argued that without any post-task change, any in-task effect would have minimal effect upon exercise adherence.

Before any efforts are made to implement modified procedures, the research of Wild et al. (2001) and Hennig-Thurau et al. (2006) should be considered. They found that emotional reactivity to facial stimuli tended to diminish in response to repeated exposure. As such, the very act of ensuring multiple viewings might actually be counterproductive. This might offer another explanation for the current outcomes. Perhaps, the EC effect was apparent in the early stages of the task and waned, due to repeated viewings, as the experimental trial progressed. To assess the veracity of this explanation, any subsequent research could consider analyzing the EC effect over the time-course of an exercise task. If the assertion holds, then emotional stimuli might only provide benefits if introduced in the latter states of an exercise session.

Regarding the actual stimuli, while the use of static facial images has been shown to be a valid transfer medium, dynamic representations could prove to be more powerful (Hatfield, Bensman, Thorton, & Rapson, 2014). For example, Hess and Blairy (2001) used dynamic images and observed the EC effect even when the emotional displays were relatively weak. Unfortunately, adopting this approach could compromise external validity. For example, gymnasia would have to invest in equipment that allowed dynamic images to be displayed. This would not be worthwhile unless the EC effect was quite strong.

This study differs in a major respect to those that have shown EC in real-world sport situations. In Totterdell (2000) and Moll et al. (2010), active participants would have had multiple sources of emotional transfer; for example, verbal and body gestures as well as facial expressions. Under the conditions of the current study, a single EC source may not have been strong enough to induce the effect. EC can occur via a number of transfer mechanisms (e.g., speech, facial expression, body gestures), so perhaps a multifaceted approach is required. Again, efforts to increase the strength of the stimuli could actually reduce the practicality of utilizing EC in this environment.

It is also possible that the EC effect did occur and was not detected. Although efforts were made to justify selection of dependent variables, there are problems with using RPE and enjoyment measures. Addressing effort sense, while a negative linear relationship between affect and RPE has been suggested, the assumption appears to be that reductions in RPE improve affect (Hardy & Rejeski, 1989). It is currently unclear whether this association is reciprocal. The assumption the RPE would be influenced by affective change might be misguided. Enjoyment was included because of the hypothesized relationship with affect (e.g., Miller et al., 2005). While many theorists assert that these two constructs are linked, others claim that enjoyment is a complex, multifaceted construct that is not necessarily synonymous with affect (Kimiecik & Harris, 1996). In retrospect, an increase in enjoyment should not necessarily have been expected. That said, Raedeke (2007) did find that enjoyment was significantly related to post-task affect, as measured via the Vigor subscale of POMS (r = .39, d = 0.84). It is also worth noting that Enjoyment decreased from Trial 1 to Trial 2, regardless of condition. Perhaps, exposure to the first trial induced a novelty effect that inflated the initial enjoyment ratings (Kimiecik & Harris, 1996; Onwuegbuzie, 2003). These arguments aside, given the importance of such variables to the exercise experience, the fact the stimuli failed to influence them would imply that the EC effect would be of little value to the exerciser.

The final discussion point relates to one of the central tenets of EC theory, namely mimicry. Many believe that facial mimicry must occur if an emotion is to be felt, and it is the afferent feedback from the facial muscles that produces the corresponding emotional experience (Hatfield et al., 2014; Howard & Gengler, 2001). Unfortunately, no assessment of facial mimicry was made, so there is no way of knowing if it occurred. This is a major limitation. Other researchers (e.g., Hess & Blairy, 2001) have addressed this issue through the use of video equipment. This has allowed the facial expressions of the receiver to be recorded and analyzed for emotional reactions. However, while determining whether mimicry has occurred is important, adopting this procedure might be problematic, particularly in the exercise environment. The introduction of such equipment in the exercise setting could provoke evaluation apprehension (performance anxiety; Elliott, Polman, & Taylor, 2012). It must also be recognized that while numerous studies do support the mimicry hypothesis (Hatfield et al., 2014), Hess and Blairy (2001) induced the EC effect without facial mimicry. Despite such concerns, future research into this area needs to make efforts to determine whether the mimicry response occurs.

In summary, this is the first attempt to induce the EC effect in the exercise environment. Had the hypothesized results been forthcoming, the simple act of displaying positive images in gymnasia might have offered a simple means of enhancing the exercise experience and ultimately adherence. Unfortunately, the effect was not observed. The authors felt that the design utilized was appropriate, as it had reasonably high internal and external validity. However, it is possible that methodological issues affected outcomes so a number of recommendations have been made to help direct any future research. It is worth restating that the authors never intended to apply any positive results to this particular group-exercise setting. During such tasks, any emotional transference would most probably come from other group members or the task leader. Rather, the exercise task was selected because it seemed most appropriate to make an initial enquiry into the application of EC in the exercise environment. While alternative approaches might help uncover an effect in exercising participants, any efforts to induce the effect must be weighed against threats to validity. Many of the recommendations made at this stage are mere conjecture. Overall, these results could suggest that the either (a) the EC effect does not manifest in this setting, (b) if present, provides no benefit to exercise-related psychological states, or (c) methodological issues hampered efforts to observe the phenomenon. Despite its failings, this investigation initiates research on the potential of EC within the exercise environment.